一维地震预报评分问题

ON THE PROBLEM OF GRADING FOR ONE-DIMENSIONAL SEISMOLOGIC PREDICTIONS

本文对当前应用较多的各种基于2×2列联表的一维地震预报评分进行了较全面地评述,文中论及: 1.从统计相关性着眼,Wallen评分V=^(1/2)QS是最合理的预报评分. 2.顾氏评分S、许氏评分R及O评分Q虽便于使用,但各有不足. 3.由V可引入一系列评分,S与Q的调和均值、Hedike评分S_H皆在此系列中.定理5-3列出了常见评分的顺序. 4.从重合程度着眼给出了评分W,并由之引出另一个标准化评分列——W系列(标准化后的有效指数h。在其中居于最“严”的位置).

This paper gives a complete commentary on some popular grading methods which are based on the 2×2 contingency table, for one-dimensional seismologic predictions. It contains:1. In view of statistical correlation, the Wallen's grading V =SQ^(1/2) is the most reasonable one among all gradings.2. Each one of the three gradings, Gou Zhen-chao's grading S =n11/n1.-n01/no. Obu-hov's grading Q=n11/n.1-n10/n.o and Xu Shaoxie's grading R =n11/n.1-n1./N, has its owninadequacy; although they are so easy to use.3. A set of gradings can be derived from V. The harmonic mean of S and Q,H(S, Q) and the Hedike's grading SH are members of this set. The order of some popular gradings are given in Theorem 5-3.4. In view of the measure of the overlapping part a grading W =n11/N-n00 isintroduced. Another set of gradings may be derived from it. The index of efficiency is the most strict one of this set.